Mail Archives: cygwin/1998/11/18/09:12:17
> >
> > Here are a few things I noticed while moving from b19 -> b20:
> >
> > Cygwin b20 Nov 2 release
> > running NT4 SP3
> >
> > 1. cpp seems to not work right. Instead of stripping out an undefined
> > #ifdef/#endif block, it strips the preprocessor lines and leaves the
> > block intact. I don't know how general this is. In particular,
> > when compiling libX11 (R6.3 patch, modified), "make X11.def" gives
> > a bogus file. Replacing the cpp call with "gcc -E -" solved the
> > problem. I noticed that this was done in the R6.4 patch.
>
> Stephen,
>
> What cpp were you using? The one supplied with gcc, which is buried way
> down in the compiler directory or some other (eg., Borland also supplies a
> cpp.exe, which causes lots of problems with configure scripts that look
> for the cpp executable).
>
> Regards,
> Mumit
>
I can't reproduce this, it seems to work OK now. The problem occurred
when compiling X11R6.3, with config/cf/cygwin32.* changed to support B20.
I had built this previously with b19. The make failed when building the
libX11 dll, specifically when generating "X11.def" from the "X11-def.cpp"
file (the Makefile invokes cpp to do this). It was as if the __CYGWIN32__
variable was not defined, leaving extra stuff in the output file. I ran
the command by hand and got the same result ("make X11.def"). The problem
went away when I replaced the cpp invocation
"gcc -print-prog-name=cpp | sed 's@\\\\@/@g'`" with
"gcc -E -". The __CYGWIN32__ was defined in the command string in both
cases. The entire build succeeded after this fix. The cygwin "cpp" is
the only cpp on that system.
Since then the major changes I remember are, switching to a binary mount
for /, and installing the cygwin1.dll from "coolview". Now when I go
back any put the original command back in the R6.3 libX11 Makefile, it
builds correctly.
Steve
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request AT cygnus DOT com" with one line of text: "help".
- Raw text -